network-slimming icon indicating copy to clipboard operation
network-slimming copied to clipboard

关于L1 regular

Open alexxiecan opened this issue 2 years ago • 1 comments

亲爱的Eric,关于m.weight.grad.data.add_(args.storch.sign(m.weight.data)),我记得损失+L1regular公式有一个求和的,请教下m.weight.grad.data.add_(args.storch.sign(m.weight.data))为什么没有求和呢?望答复,谢谢!

alexxiecan avatar Nov 16 '22 02:11 alexxiecan

论文中的公式(1)是有求和的

alexxiecan avatar Nov 16 '22 02:11 alexxiecan